A composite image generation system comprises: a position information detection device 14 that detects position information about a predetermined spot in a real space; and a processing device 16 that generates a composite image in which a separately acquired superimposed image is overlaid onto a desired region in a camera image of the real space captured by a movably placed video camera 18, based on the position information from the position information detection device 14. The processing device 16 includes a position information specifying means 35 that adjusts a detection result from the position information detection device 14 to specify position information about a probe 19 used when superimposing the superimposed image onto the desired region. The position information specifying means 35 comprises a recognition state determination unit 46 that determines a recognition state of a probe marker 24A by the video camera 18, a used information selection unit 47 that selects which of detection results from an optical tracking means 21 and a non-optical tracking means 22 is to use according to the recognition state, and an output position information decision unit 48 that decides output position information necessary to generate a composite image according to the selected position information.